本文选取32个小分子化学成分和9 个天然产物化学成分为训练集，利用Dragon软件计算分子描述符，应用WEKA 软件中CfsSubsetEval评价方法和BestFirst-D1-N5搜索方法进一步筛选描述符，利用Linear Regression方法建立了分子结构与其血清蛋白结合常数值logK 的定量结构-性质相关性模型（相关系数为0.9557）。应用该模型对10个绿原酸类化合物的logK进行预测，结果显示绿原酸的logK预测值为4.51，与文献报道的logK值在4-7范围内基本相符。研究结果为进一步探讨绿原酸与血清蛋白的结合情况提供指导，为研究药物与血清蛋白的结合提供思路。

Molecular descriptors of the training set, which include 32 small molecules and 9 natural compounds, were calculated by the Dragon software. CfsSubsetEval evaluation methodology and BestFirst -D1 -N5 searching method were further used to screen molecular descriptors. Quantitative structure -properties relationship (QSPR) model was established by linear regression method to analyze the relationship between molecular structure and the binding constant logK of serum albumin. The model, with the correlation coefficient of 0.9557, was used to predict logK value of 10 chlorogenic acids. The predictive logK of chlorogenic acid is 4.51, which is consistent with the value 4-7 that reported in literatures. The results will be useful for further study on the combination of chlorogenic acid and serum albumin.

北京中医药大学自主课题（2011-CXTD-11）：中药信息工程，负责人：乔延江；北京中医药大学自主课题（2011-JDJS-12）：活血化瘀中药治疗冠心病的作用机制研究，负责人：张燕玲；教育部博士点基金（20092213120006）：基于花生四烯酸代谢途径的清热解毒中药有效成分信息辨识研究，负责人：张燕玲。